Culture Model of Human Corneal Epithelium for Prediction of Ocular Drug Absorption

Abstract

purpose. The main purpose of this study was to develop a cell culture model of immortalized epithelium from the human cornea for drug permeability testing.

methods. Immortalized human corneal epithelial (HCE) cells were grown on filters, with various filter materials and coating procedures. In the optimal case, HCE cells were grown on polyester filters coated with rat tail collagen gel containing fibroblast cells. Transepithelial electrical resistance (TER) was measured during the growth of the cells to evaluate the epithelial differentiation and tightness of the epithelial cell layers. Transmission electron microscopy (TEM) was used to show the formation of tight junctions, desmosomes, and microvilli. Cellular morphology was characterized by light microscopy. Permeabilities of ³H-mannitol and 6-carboxyfluorescein were determined, to evaluate the intercellular spaces of the epithelium. Rhodamine B was used as a lipophilic marker of transcellular permeability. Permeabilities of the excised rabbit corneas were determined in side-by-side diffusion chambers.

results. The TER values of the corneal epithelial cultures were 200 to 800Ω · cm², depending on the culture conditions. In optimal conditions, cultured corneal epithelium consisted of five to eight cell layers, TER was at least 400 Ω · cm², and the most apical cells were flat, with tight junctions, microvilli, and desmosomes. The permeability coefficients (P _cell, 10⁻⁶ cm/sec) for ³H-mannitol, 6-carboxyfluorescein, and rhodamine B were 1.42 ± 0.36, 0.77 ± 0.40, and 16.3 ± 4.0, respectively. Corresponding values (at 10⁻⁶ cm/sec) for the isolated rabbit corneas were 0.38 ± 0.16, 0.46 ± 0.27, and 18.1 ± 4.0, respectively.

conclusions. The TER, morphology, and permeability of the cultured corneal epithelial cells resemble those of the intact cornea. This cell culture model may be useful in evaluation of corneal drug permeation and its mechanisms.

After topical instillation, drugs are absorbed into the inner eye through the cornea or the conjunctiva and sclera. The cornea is the main route of absorption for clinically used ocular drugs. ^{1 2 3} Permeability studies of corneal drugs are usually performed in vitro using isolated rabbit corneas mounted in modified Ussing chambers. ^{4 5 6} In these studies, numerous rabbits are needed and the isolated corneas are viable for only 6 hours after dissection. In addition, differences between species may impair the predictability of the results in terms of drug absorption in humans. 

In vivo studies of ocular drug absorption usually require killing at least five animals at each time point in the drug concentration profile. Because full pharmacokinetic analysis includes many time points, the number of rabbits used per study is often more than 20. ⁷ Although rabbits are the only widely used animal model in ocular pharmacokinetics, this model has several problems, among which is the rabbit’s low blinking frequency compared with the human’s. ⁸  

In general, cell culture models offer many potential advantages in the analysis of drug transport and metabolism. ⁹ Of particular importance, they open possibilities to decrease the number of animal experiments in this type of research. In addition, these systems offer potential for manipulation of the environment or cellular properties as a means of addressing the mechanistic questions of drug permeation in living cells. From the standpoint of drug discovery and drug formulation, cell culture models can be used to expedite identification of compounds or formulations with favorable pharmacokinetic properties and to evaluate structure–absorption and structure–metabolism relationships on a large scale. Development of new test methods of in vitro absorption is of utmost importance in current drug discovery, because the modern combinatorial and automated methods of drug synthesis produce numerous compounds, but only in minute quantities. Therefore, maximal predictive information should be gained from the in vitro use of small quantities of the test substances. 

A corneal cell culture model could be useful in testing the permeability of ocular drugs and formulations. The model should be based on corneal epithelial cells, because the epithelium is the barrier that limits permeation. ¹⁰ The published models are based on primary cells, ^{11 12 13} but these cells usually stop growing after one or two passages and after storage in liquid nitrogen, they revive weakly or not at all. New rabbit cells must be isolated frequently; therefore, these methods may not be optimal for larger scale screening of new compounds and formulations. 

Immortalized cell lines can be grown continuously, and they should be more practical for testing of permeability. SIRC cells are a corneal cell line from rabbits that was recently described for permeability testing. ^{14 15} Unfortunately, these cells exhibit a fibroblast phenotype, which decreases their value as a model. ¹⁶ The human corneal epithelial (HCE)-T in vitro model of human corneal epithelium ^{17 18 19} represents a three-dimensional culture for HCE-T cells grown on a collagen membrane to provide a species- and tissue-specific equivalent of the human corneal surface in vivo. Araki-Sasaki et al. ²⁰ established an immortalized HCE cell line that has properties of normal corneal epithelial cells. This cell line continues to grow for more than 400 generations through 100 passages, and the cells can be frozen and revived. Recently, Saarinen-Savolainen et al. ²¹ demonstrated the usefulness of this cell line in testing the toxicity of ophthalmic drugs and pharmaceutical excipients, but so far no permeability model based on immortalized HCE cells has been presented. 

The purpose of this study was to establish an in vitro corneal permeability model using an immortalized HCE cell line. We suggest that this model may be useful as a predictive kinetic model in vitro, because the cultured corneal epithelium exhibits morphologic characteristics and permeability similar to those observed in the intact cornea. 

Materials and Methods

Cell Culture

Immortalization of HCE cells has been described earlier. ²⁰ Polyester and polycarbonate cell culture filters (surface area, 4.7 cm²; pore size, 0.4 and 3.0 μm; Transwell Clear; Transwell Costar, Cambridge, MA) were used with no coating or were coated with 275 μl rat tail collagen type I (1.3 mg/ml; Becton Dickinson, Bedford, MA). Collagen was mixed with 20,000 mouse fibroblasts per milliliter (BALB 3T3) and allowed to gel on the filters at room temperature for at least 4 hours. Suspension of mycoplasma-free HCE cells, passages 22 to 35, were seeded onto the coated filters at a concentration of 90,000 cells/cm². The cells were grown at 37°C in humidified air with 5% CO₂, in standard culture medium both in apical and basolateral chambers for 7 to 10 days until the cells were confluent. The cells were then exposed to an air–liquid interface for 2 to 3 weeks. Filters coated only with collagen or collagen-laminin (Becton Dickinson) and filters without any coating were also used. The culture medium was replaced every other day. The standard medium consisted of DMEM/Ham’s F12 (1:1) (Gibco BRL, Grand Island, NY), 15% heat-inactivated fetal bovine serum (FBS; (Gibco BRL), 0.3 mg/ml l-glutamine (Gibco BRL), 5 μg/ml insulin (Sigma), 0.1 mg/ml cholera toxin (Calbiochem, La Jolla, CA), 10 ng/ml EGF (Calbiochem), 0.5% dimethyl sulfoxide (DMSO; Sigma, St. Louis, MO), 0.1 mg/ml streptomycin, and 1000 IU/ml penicillin (both from Gibco BRL). 

Transepithelial Electrical Resistance

Transepithelial electrical resistance (TER) was metered (Endohm; World Precision Instruments, Sarasota, FL) at different phases of cell growth. TER was used as an indicator of epithelial differentiation and epithelial tightness. TER data were corrected for low-background TER by using a blank filter containing the possible coating materials and culture medium. The TERs of the filter without HCE cells were 47Ω · cm² on polyester filters and 28Ω · cm² on polycarbonate filters. TERs of blank filters were not dependent on filter coating. At the end of each permeability experiment, TER was measured to detect the condition of the cells. The TER data were not changed during the permeability studies. 

Permeation Studies with Cultured Cells

The permeation studies were performed by using ³H-mannitol (molecular weight [MW], 182; specific activity, 19.7 Ci/mmol; NEN Life Science Products, Inc. Boston, MA) and 6-carboxyfluorescein (MW 376; Sigma) as hydrophilic markers for characterizing the paracellular permeation between the epithelial cells. The experiments were performed after cell growth in different conditions. The penetration study with ³H-mannitol was initiated by adding 2.6 ml serum-free medium to the basolateral side (receiver side) and 1.5 ml serum-free feeding medium containing ³H-mannitol (0.5 μCi/ml) to the apical side (donor side). At 30, 60, 90, 120, 150, 180, 240, and 300 minutes, aliquots of 100 μl were withdrawn from the receiver chamber and replaced with an equal volume of blank medium. Then 400 μl scintillation liquid (Optiphase Supermix; Milton Keynes, Wallac, UK) was added to every sample, and the radioactivities were measured using a liquid scintillation and luminescence counter (1450 MicroBeta Trilux; Wallac, Turku, Finland). 

The transport was studied also with 6-carboxyfluorescein (Sigma) as a probe, in a manner similar to the ³H-mannitol experiments, except that balanced salt solution (BSS; BSS Plus; Alcon, Fort Worth, TX) was used as a buffer solution and the concentration of 6-carboxyfluorescein on the apical side was 10 μM. The salt solution was used because the red color of the medium could disturb the fluorescence measurements. The sample volume was 200 μl. Concentrations of 6-carboxyfluorescein were determined using a fluorescence plate reader (model FL 500; Bio-Tek Instruments, Inc., Burlington, VT) with 485-nm excitation and 530-nm emission filters. 

The lipophilic marker was rhodamine B (MW 479; Sigma) in BSS. The donor concentration of rhodamine B was 10 μM, and samples were taken at 10, 20, 30, 45, 60, 75, 90, 105, and 120 minutes. Concentrations of rhodamine B were determined by the fluorescence plate reader with 530-nm excitation and 590-nm emission filters. To increase the fluorescence intensity, methanol was added into the samples (the final concentration of methanol was 40% vol/vol). Otherwise, the procedure was identical with the 6-carboxyfluorescein experiments. To avoid buildup of the limiting hydrodynamic diffusion layers all permeation studies were performed at 37°C using a horizontal plate mixer (Heidolph Inkubator 1000 and Titramax 1000; Heidolph Electro GmbH& Co., Kelheim, Germany). 

Light Microscopy

Confluent corneal epithelial cells were fixed in 2% paraformaldehyde-0.5% glutaraldehyde phosphate buffer, dehydrated in a graded ethanol series (80%, 94%, 99%) and embedded in paraffin. Sections of 5 μm were cut and stained in hematoxylin and eosin solutions. The specimens were examined with a light microscope (Microphot-Fxa; Nikon, Tokyo, Japan). 

Male albino New Zealand strain rabbits, weighing approximately 3 kg, were used to study the morphology of normal rabbit cornea. The corneas were fixed in 4% buffered formalin solution, dehydrated in a graded ethanol series, embedded in paraffin, cut in 5-μm sections, and stained with hematoxylin-eosin for light microscopy. 

Transmission Electron Microscopy

The cultured cells were fixed in 2.5% glutaraldehyde, postfixed with 1% osmium tetroxide, dehydrated in graded alcohols, embedded in resin (LX-112; Ladd Research Industries, Inc., Burlington, VT) and cut in 60-nm sections. The specimens were viewed and photographed on a transmission electron microscope (JEM 1200 EX; JEOL, Tokyo, Japan). 

Permeability Experiment with Excised Cornea

All animal experiments conformed to the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. Male and female New Zealand albino rabbits (age 6–12 months, weight 3.0–4.6 kg) were killed by injecting a lethal dose of fentanyl and fluanisone (Hypnorm; Janssen Pharmaceutica, Beerse, Belgium) into the marginal ear vein. The cornea was dissected with a scleral ring (5 mm laterally from the limbus). The cornea was attached to a corneal holder with a rubber band (made in our laboratory) using a tissue mounting tool (Physiologic Instruments; San Diego, CA), and the lens and iris were carefully removed. The corneal holder with cornea was mounted between two side-by-side diffusion chambers (Snapwell; Navicyte, Sparks, NV). The exposed corneal surface area was 0.6 cm². Blank BSS buffer (6.5 ml) was added to the endothelial side and the donor solution (6-carboxyfluorescein and rhodamine B, both at a concentration of 100 μM, dissolved in 5.5 ml BSS buffer) was added to the epithelial side. The solutions in each chamber were mixed by bubbling an O₂-CO₂ (95:5) mixture at the rate of three to four bubbles per second and maintaining the temperature at 35 ± 1°C in a circulating water bath. At intervals of 30 minutes, until 240 minutes, samples of 500 μl were taken from the endothelial side and replaced by an equal volume of blank BSS buffer. The 6-carboxyfluorescein and rhodamine B were analyzed, as described earlier. Permeability of ³H-mannitol in isolated corneas was studied, as described by Suhonen et al. ²²  

Data and Statistical Analysis

Apparent permeability coefficients (P _tot, in centimeters per second) of the cultured HCE cells and filter together were calculated as  

\[P_{\mathrm{tot}}{=}V(dc/dt)/AC_{\mathrm{0}},\]

where P _tot is the apparent permeability of the filter and cells together, dc/dt is the drug flux across the membrane, V (in cubic centimeters) is the volume in the receiver side, A (in square centimeters) is the surface area of the filter, and C ₀ (in millimolar) is the initial concentration in the donor compartment. 

The cultured corneal epithelium and the support (filter, extracellular matrix, feeder fibroblasts) caused serial resistance against drug permeation. The resistance for drug permeability was provided both by the barrier of the support (R _sup) and cell layer (R _cell). Total resistance of the system (R _tot) was defined as  

\[R_{\mathrm{tot}}{=}R_{\mathrm{cell}}{+}R_{\mathrm{sup}}\]

Permeability is the inverse value of the barrier resistance: R = 1/P. Thus  

\[1/P_{\mathrm{tot}}{=}1/P_{\mathrm{cell}}{+}1/P_{\mathrm{sup}}\]

Permeability of the cultured corneal epithelium without support ¹² is obtained by rearranging equation 3 to the following form:  

\[P_{\mathrm{cell}}{=}P_{\mathrm{tot}}{\times}P_{\mathrm{sup}}/(P_{\mathrm{sup}}-P_{\mathrm{tot}})\]

where P _tot is the apparent permeability of the filter and the cells together (obtained experimentally using equation 1 ), P _sup is the apparent permeability of the filter and possible coatings without the cells (obtained experimentally), and P _cell is the apparent permeability coefficient of the cell layer calculated by equation 4 . P _sup was always approximately 30 × 10⁻⁶ cm/sec and was not dependent on the filter coatings. In this system the major limiting layer of permeation was always the HCE cells. Because permeabilities of the isolated corneas are determined without any filters or coatings, P _cell is the most logical value for comparisons. 

The Mann-Whitney test was used to study the statistical significance of the differences between the permeabilities of model compounds in the isolated rabbit cornea and cultured HCE cells. P < 0.05 was considered to show a statistically significant difference. 

Results

TER and Permeability

Without the air–liquid interface, the TER values of the cultured corneal epithelia were approximately 100 to 200Ω · cm² (Fig. 1) , and the P _cell of ³H-mannitol was 10 to 20 × 10⁻⁶ cm/sec, regardless of the filter or coating material (data not shown). Under the air–liquid interface the transepithelial TER of the cultured human corneal epithelium increased for 2 to 3 weeks to approximately 200 to 800Ω · cm², depending on the culture conditions (Fig. 1) . Permeability was determined when the TER was at least 85% of peak level. The time when the resistance reached a constant level varied from 3.5 to 5 weeks after seeding, depending on the coating material of the filters (Fig. 1) . Figure 2 shows the relationship between TER and P _cell, for ³H-mannitol in different cases. With decreasing TER, the permeability coefficient was increased. At a TER of 400 to 800Ω · cm², P _cell was 1 to 2 × 10⁻⁶ cm/sec. TER was similar before the permeability experiment and after the experiment. 

The pore size of the filters was an important factor. Regardless of the coating method, the cells migrated through the filter with 3-μm pore size. For this reason, these filters were not a suitable support for the cultures. In each coating system, the polyester filters were better than polycarbonate filters for culturing the HCE barrier (Fig. 2) . Polyester filters (0.4 μm) coated with collagen or collagen-fibroblasts or collagen-laminin were the best system for culturing HCE cell layers. In these conditions the P _cell for mannitol was 1 to 2 × 10⁻⁶ cm/sec, and the corresponding TER was approximately 400 to 800 Ω · cm² (Fig. 2) . Therefore, the permeabilities of 6-carboxyfluorescein (paracellular permeant) and rhodamine B (transcellular permeant) were studied using polyester filters coated with collagen-containing fibroblasts. In cultured corneal epithelium (with collagen and fibroblasts on polyester), rhodamine B penetrated the cell layers 21 and 11 times faster than hydrophilic 6-carboxyfluorescein or ³H-mannitol, respectively. In excised rabbit cornea, the differences were 39- and 48-fold, respectively. The HCE cell culture model was more permeable to ³H-mannitol than isolated rabbit cornea (P < 0.05), whereas in the case of rhodamine B and 6-carboxyfluorescein, no significant differences were seen (Table 1) . 

Light and Transmission Electron Microscopy

Figure 3A shows the morphology of rabbit corneal epithelium. The normal epithelium is divided into two to three layers of flattened superficial cells, two to three layers of polygonal cells, and a single cell layer of columnar basal cells. A micrograph of histologic cross section of cultured HCE cells is shown in Figure 3B . Light microscopy and TEM experiments did not show any difference in morphology between the three culture methods: the corneal epithelial cells on polyester filters with collagen and fibroblasts, collagen, or collagen with laminin. The epithelium consisted of five to eight cell layers (Fig. 3B) . The most apical cells were flat (Fig. 3B) with tight junctions, microvilli, and desmosomes (TEM; Fig. 4 ). Thickness of the cultured corneal epithelium was 70 μm, which is close to the thickness of the human corneal epithelium (50–70μ m). ²³ Rabbit corneal epithelium is seen in Figure 3A . Its thickness is approximately 40 μm. When the cells were grown on polycarbonate filters, 3 to 10 cell layers were formed without flattening of the apical cell layers (Fig. 5) . However, there were some tight junctions, microvilli, and desmosomes. If the cells were cultivated without the air–liquid interface, only two to three cell layers without flattened cells and tight junctions were formed (Fig. 6) . 

Discussion

After topical ocular application, drugs may be absorbed into the eye through the corneal or conjunctival and scleral route. ^{1 2 3} The route through conjunctiva and sclera is important mostly for very hydrophilic and large molecules that are not able to penetrate through the corneal barrier. ²⁴ Most clinically used ocular drugs have adequate lipophilicity for corneal absorption, and such properties are sought in the development of new ocular drugs. For these reasons, a viable permeability screening system for corneal absorption is useful. Conjunctiva and sclera are considered to be an important alternative route for high-molecular-weight drugs (e.g., proteins) and specialized systems for drug delivery to the posterior segment. ^{25 26 27}  

In this study, a model of drug permeability was developed using immortalized epithelial cells from human cornea. Three earlier studies have described cell culture models that use primary epithelial cells from rabbit cornea. ^{11 12 13} However, immortalized cells have practical advantages such as revival after freezing, no need for frequent and tedious cell isolation, and use of fewer rabbits. Such a model should resemble an in vivo cornea as closely as possible. We used immortalized epithelial cells from human cornea, because the epithelium is the main barrier to drug absorption. ¹⁰ Furthermore, use of human cells is advantageous in a cell culture model, because humans and rabbits may differ significantly in metabolic enzymes and active transporters. Such cell culture models are, however, dependent on optimal cell culture conditions. 

In our study, the feeding medium of the cells included 15% FBS. Sometimes, however, high concentrations of FBS may disturb cell proliferation and differentiation, ²⁸ and some groups have reported improved differentiation of the corneal epithelium in serum-free medium. ^{19 29} We also cultivated the cells in lower FBS concentrations (2%, 5%, and 10%), but in these conditions the cells did not survive (data not shown). Thus, the serum concentration was maintained at 15%, and the other medium components were similar to those described by Araki-Sasaki et al. ²⁰ As described for the primary corneal epithelial cells, ³⁰ the air–liquid interface was critical for the differentiation of HCE cells in culture. No flat apical cells or proper barrier was obtained without air lifting. 

Other important factors in development of the culture model are the filter material, its pore size, and the coating components on the filter (i.e., the extracellular matrix). Polycarbonate, with or without collagen, is used in most cultures of corneal cells. ^{12 13 31} For barrier formation with HCE cells, however, polyester filters were better than polycarbonate filters (Fig. 2) , and the clear polyester membrane also provided better cell visibility for light or phase–contrast microscopy. 

Collagen facilitates the growth and differentiation of corneal epithelial cells in culture. ^{12 30 31 32 33 34} Collagen coating helps the cells to attach to the cultivating bed and stimulates their proliferation and differentiation. Human corneal epithelium cultured on collagen gels can synthesize and deposit basement membrane components such as laminin and type IV collagen. ^{34 35} The mixture of collagen and laminin further improved the culture model and in our studies represented the basal lamina (Fig. 2) . 

3T3 fibroblast cells from mice are used as a feeder layer for differentiation of the epithelial cells. The interactions between the fibroblasts and corneal epithelial cells stimulate the differentiation of the corneal epithelial primary cells. ^{31 36} For culturing immortalized HCE cells, we coated the filters with a mixture of collagen and fibroblasts to provide a substrate resembling a corneal stroma. The fibroblasts did not divide in the collagen matrix, but they functioned as feeder cells for HCE cells. Based on morphology and on the barrier properties, a polyester filter with collagen and fibroblasts appeared to be the best conditions for HCE cell culture. 

The apical surface of the corneal epithelium contributes more than half of the total electrical resistance of the cornea, ³⁷ and the top two layers are the most important part of the cornea in limiting the permeability of hydrophilic drugs. ^{2 10} Therefore, the most apical cell layers are the most important part of the HCE permeability model. Of importance, the HCE cell culture model shows tight junctions and desmosomes in the flattened apical cell layer. Although the apical part of the cultured corneal epithelium appears to include the major structural features, it is also noted that the wing cells and basal cells in the cultured epithelium are not organized as well as they are in the intact cornea (Fig. 3) . Obviously, the HCE permeability model does not include stroma or endothelium, but these layers are not critical barriers to corneal drug absorption. 

In optimal conditions, the TERs of the HCE cell cultures were 400 to 800 Ω · cm². This is approximately 10 times lower than the TER of 3 to 10 kΩ · cm² in the excised rabbit cornea. ^{38 39 40} In addition, the permeability of the HCE cell cultures for hydrophilic markers mannitol and 6-carboxyfluorescein was two to four times higher than in the isolated rabbit cornea (Table 1) , although the difference for 6-carboxyfluorescein was not statistically significant. These markers have a very low lipophilicity, and they permeate biological membranes only through the intercellular space. The higher permeability and lower TER of the HCE cell cultures suggest that the cell culture exhibits a slightly larger intercellular space (i.e., the pore size or pore density is larger) compared with rabbit cornea. Pore size and pore density will be evaluated in a quantitative manner in future studies. 

Nevertheless, permeability of the HCE cell cultures for hydrophilic markers (0.8–1.4 × 10⁻⁶ cm/sec; Table 1 ) is low enough to allow the screening of drug candidates. Ophthalmic drugs are usually administered as eye drops, and their sites of action are often in the inner eye structures. These drugs are small and lipophilic enough for transcellular permeation. They exhibit corneal permeability between 5 and 30 × 10⁻⁶ cm/sec (e.g., betaxolol, dexamethasone, timolol), ⁴¹ which is 4 to 20 times higher than the paracellular permeability of mannitol and carboxyfluorescein in the HCE cell culture model. To verify that the HCE cell culture is useful in screening, the permeability of a lipophilic transcellular marker rhodamine B was determined. As expected, rhodamine B exhibited high permeability of the HCE cell culture (16 × 10⁻⁶ cm/sec), practically the same as in the isolated rabbit cornea (Table 1) . Furthermore, the permeabilities of 6-carboxyfluorescein and rhodamine B are in the same range with the previously reported values for these compounds in the excised rabbit cornea. ⁴²  

The difference in permeability of rhodamine B and 6-carboxyfluorescein was 21-fold, although not quite as large as in the isolated rabbit cornea (39-fold). In contrast to our model, Kawazu et al. ¹² reached only a 4.6-fold difference between hydrophilic atenolol (P _cell 2.6 × 10⁻⁵ cm/sec) and lipophilic alprenolol (10.8 × 10⁻⁵ cm/sec) using primary corneal epithelial cell cultures. Our results indicate that the HCE cell culture model can discriminate between the permeabilities of lipophilic and hydrophilic drugs. In addition, comparison of the permeabilities in the HCE cell culture model (Table 1) with a database of more than 100 compounds ⁴¹ shows that the values for both hydrophilic drug (6-carboxyfluorescein) and lipophilic drug (rhodamine B) match the average corneal permeability for the drugs with similar hydrophilicity-lipophilicity very well. 

Only sparse information is available about the permeability of the human cornea for drugs. The rabbit corneas and rabbits in vivo are the current standard methods in ocular drug delivery studies. Therefore, we compare the cell culture model to the rabbit cornea instead of the human cornea. The morphology of the human and rabbit cornea are similar: The main permeability barrier in both cases is in the apical tight layer of the corneal epithelium. Human cornea and its epithelium (0.52 mm and 50–70 μm) are thicker than the rabbit cornea (0.35–0.42 mm) and its epithelium (40 μm). Permeabilities of cyclophosphamide and four carbonic anhydrase inhibitors have been compared in human and rabbit cornea. ^{8 43 44} The permeabilities are similar, but the paracellular permeability in rabbit cornea may be less than in human cornea. ^{8 43 44} This type of difference was also seen in this study: The paracellular space of the HCE culture model appeared wider than in the rabbit cornea. The existing morphologic and permeability data suggest that the rabbit cornea is a fairly good model. Studies with cultured human corneal epithelial cells will provide further information about the similarities and, importantly, also about the differences between the human and rabbit corneal epithelial transport processes relevant to drug absorption. 

It appears that the permeability barrier of the HCE culture is comparable to that of rabbit cornea. Therefore, this model can be used to predict ocular absorption of the drugs that permeate through the cornea by passive diffusion. The extent of applicability of this model also depends on the expression of the active transporters, efflux proteins, metabolic enzymes, surface proteins, and mucins in the cultured cells. These aspects of the cultured model should be elucidated in future studies, because it is possible that the expression profiles of the immortalized cell line, primary cells, and animal model in vivo are different. An in vitro cell culture model does not take into account the issue of contact time on the corneal surface. Before an experimental setup is developed for such cell culture studies, the impact of pharmaceutical formulations with prolonged precorneal contact should be investigated in vivo in animal studies. 

In conclusion, we have developed an in vitro model of the corneal epithelium using immortalized human cells. This model resembles intact cornea with morphologically identifiable desmosomes, tight junctions, microvilli, and cell layers with apical flat cells. The discrimination capacity of the hydrophilic and lipophilic markers and their overall permeabilities in the cell culture model suggest that this model can be used to evaluate corneal drug permeabilities and the mechanisms of ocular drug absorption. 

 Supported by the Academy of Finland and the Julia von Wendt Foundation.

 Submitted for publication April 11, 2001; revised July 2, 2001; accepted July 30, 2001.

 Commercial relationships policy: N.

 The publication costs of this article were defrayed in part by page charge payment. This article must therefore be marked“ advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.

 

Figure 1.

 

Effect of culture time of the immortalized HCE cells on TER. The immortalized HCE cells were cultured on polyester filters without any other coating material (▴), with type 1 collagen gel coating (▪), with type 1 collagen gel coating containing fibroblast cells (•), with type 1 collagen gel coating containing laminin (♦), and with collagen coating without air–liquid interface (⋄). The cells were cultured also on polycarbonate filters coated with collagen (□) or collagen and fibroblasts (○). The data in the figure are of representative individual experiments.

Figure 2.

 

The TER and the permeability coefficient (P _cell) of ³H-mannitol across HCE-cell layers (•, n = 10; all others, n = 3–5).

Table 1.

 

Permeability Coefficients P_app for ³H-Mannitol, 6-Carboxyfluorescein, and Rhodamine B

Compound	MW	Log P (oct./pH 7.4 buffer)*	P app × 10−6 (cm/sec)
						HCE Cell Culture Model	Isolated Rabbit Cornea
3H-mannitol	182	−2.2	1.42 ± 0.36, † (n = 10)	0.38± 0.16 (n = 5)
6-Carboxyfluorescein (CF)	376	−3.1	0.77± 0.40 (n = 6)	0.46± 0.27 (n = 6)
Rhodamine B (R)	479	2.3	16.3± 4.0 (n = 8)	18.1± 4.0 (n = 4)
Permeability Ratio (PR/PCF), ‡			21	39

 Permeability data are mean centimeters per second ± SD.

*  Partition coefficients between 1-octanol and buffer solution (pH 7.4).

^†  The permeability coefficient for ³H-mannitol in the HCE cell culture model is significantly larger than that in isolated rabbit cornea (Mann-Whitney test, P < 0.05).

^‡  Permeability ratio between rhodamine B and 6-carboxyfluorescein.

Figure 3.

 

Comparative light micrographs of histologic cross sections of rabbit corneal surface and immortalized HCE cells grown on a polyester filter coated with type 1 collagen gel containing fibroblast cells and stained with hematoxylin- eosin. (A) A cross section of rabbit cornea showing the corneal epithelium containing multiple cell layers and the keratocytes containing stroma. (B) A cross section of the immortalized HCE model showing the multilayer culture (approximately seven cell layers; apical side up). Scale bar, 50μ m.

Figure 4.

 

Transmission electron micrograph of immortalized human corneal epithelial cells on type 1 collagen containing fibroblasts. Narrow arrows: microvillar structures; thick arrow: a tight junction; broken arrow: a desmosome. Scale bar, 0.1 μm.

Figure 5.

 

Light micrograph of immortalized HCE cells grown on a polycarbonate filter coated with type 1 collagen gel containing fibroblast cells (apical side up). Scale bar, 50 μm.

Figure 6.

 

Light micrograph of immortalized HCE cells grown on a polyester filter coated with type 1 collagen gel containing fibroblast cells (apical side up). These cells were cultured without the air–liquid interface. Scale bar = 50 μm.

The authors thank Kaoru Araki-Sasaki, Toyanaka Municipal Hospital, Osaka, Japan, and Hitoshi Watanabe, Department of Ophthalmology, Osaka University Medical School, Japan, for kindly providing the HCE cells, and Seppo Rönkkö for advice and assistance.